基于遗传算法SVM的海洋环境腐蚀速率预测

Prediction of Corrosion Rate in Marine Environment Based on Support Vector Machine Combining with Genetic Algorithm

建立海洋环境腐蚀速率预测模型,为深海海底管道腐蚀失效风险预测提供参考,构建管道腐蚀失效预警体系。以海洋挂片腐蚀速率测量试验数据作为支持向量机(Support Vector Machine,SVM)的训练样本,SVM的关键参数经过遗传算法优化,从而解决BP神经网络等机器学习方法需要大量训练样本的问题。基于遗传算法SVM的海洋环境腐蚀速率预测模型,与人工神经网络腐蚀速率预测模型对比,可以有效降低预测误差,为深海海底管道腐蚀失效风险预测提供更为可靠的腐蚀缺陷几何参数数据,为海底管道风险预警体系的建立提供准确的数据参考依据。通过实例计算与试验数据的对比验证,证明该方法的适用性和可靠性。

The prediction model of steel corrosion rate in seawater is established to provide reference for risk assessment of deep sea pipeline corrosion failure,thus building the warning system of pipeline failure. Support Vector Machine （SVM） model is built for corrosion rate prediction. The optimal parameters are found out by genetic algorithm. The model of corrosion rate of prediction based on the GA-SVM can effectively decrease the prediction error compared to the artificial neural network model,which can offer geometric parameters of the corrosion defect and the reliable data reference for the risk warning system of pipeline. By example calculation and data comparison, the applicability and reliability of this model are proved.